# ComfyUI-AnimateDiff-Evolved: Unlimited-Length Animation with Sliding Context Windows > ComfyUI-AnimateDiff-Evolved is a ComfyUI custom node pack that brings AnimateDiff motion modules into the ComfyUI node graph, with additions including infinite animation length via sliding context windows, Motion LoRA support, FreeInit, and compatibility with HotshotXL, AnimateLCM, and AnimateDiff-SDXL. It works with nearly any KSampler node. **Kosinkadink/ComfyUI-AnimateDiff-Evolved** — Improved AnimateDiff for ComfyUI and Advanced Sampling Support - Repository: https://github.com/Kosinkadink/ComfyUI-AnimateDiff-Evolved - Stars: 3,551 · Forks: 288 - Language: Python - License: Apache-2.0 - Published: 2026-09-23 · Updated: 2026-09-23 · Language: en - Canonical page: https://hysenlabs.com/projects/kosinkadink-comfyui-animatediff-evolved ## What AnimateDiff Evolved Does and Who Needs It AnimateDiff is a method for adding temporal motion to Stable Diffusion image generation checkpoints by inserting a motion module that applies learned motion patterns across a sequence of frames. The standard AnimateDiff integration in A1111 and similar frontends handles fixed-length clips. ComfyUI-AnimateDiff-Evolved brings that capability into ComfyUI's node-based workflow system and extends it with context windowing, advanced sampling options, and support for newer motion module architectures. The target users are ComfyUI practitioners who want to generate looping video clips, animate still images, or build complex multi-stage animation pipelines using ComfyUI's composable node graph. The README states that it is compatible with almost any vanilla or custom KSampler node, meaning existing ComfyUI image generation workflows can be extended for animation without rebuilding from scratch. ## Installation and Motion Module Setup Installation via ComfyUI Manager is the recommended path: search for AnimateDiff Evolved and confirm the author is Kosinkadink. For manual installation, clone the repository into ComfyUI's custom_nodes folder. Motion modules are required and are not included in the repository. The README lists the original v1, v2, and v3 SD 1.5 motion modules (mm_sd_v14, mm_sd_v15, mm_sd_v15_v2, v3_sd15_mm) from the guoyww/animatediff HuggingFace repository, Google Drive, and CivitAI model 108836. At least one motion module is needed. Different modules produce different motion characteristics. The README describes stabilized finetunes (mm-Stabilized_mid, mm-Stabilized_high) by manshoety and a higher resolution finetune (temporaldiff-v1-animatediff) by CiaraRowles. FP16 safetensor versions are available from the continue-revolution account on HuggingFace to reduce storage while using the same VRAM. Once downloaded, motion module files go in one of two locations: ``` ComfyUI/custom_nodes/ComfyUI-AnimateDiff-Evolved/models ComfyUI/models/animatediff_models ``` Custom paths can be configured through ComfyUI's extra_model_paths.yaml file using the key animatediff_models for motion modules and animatediff_motion_lora for Motion LoRAs. ## Context Options and Infinite Animation Length Standard AnimateDiff is limited to short clips because the motion module was trained on fixed-length sequences, typically 16 frames. Running more frames directly causes quality degradation. ComfyUI-AnimateDiff-Evolved addresses this with Context Options, a sliding window mechanism that applies the motion module to overlapping windows of frames and assembles them into a continuous sequence. Context Options work at the whole UNet level, while the related View Options mechanism operates within the motion module. Both can be scheduled to change at specific points in the sampling process, which allows different motion parameters for different temporal regions of a long animation. FreeInit is available under the iteration options. FreeNoise is accessible through the noise_type dropdown in SampleSettings nodes. Both are techniques for improving temporal consistency, and both are documented in the original AnimateDiff repository and wiki rather than in this extension's README. ## Motion LoRAs, Prompt Travel, and Multival Controls Motion LoRAs modulate the motion characteristics applied by v2-based motion modules. The original LoRAs from the AnimateDiff repository work specifically with v2-based models like mm_sd_v15_v2, mm-p_0.5.pth, and mm-p_0.75.pth. A newer training approach via the AnimateDiff-MotionDirector repository produces LoRAs without that v2 limitation. Motion LoRA files go in one of two locations: ``` ComfyUI/custom_nodes/ComfyUI-AnimateDiff-Evolved/motion_lora ComfyUI/models/animatediff_motion_lora ``` Prompt travel, which changes the text prompt across frames to guide motion through content changes, is supported through built-in Prompt Scheduling nodes or through the BatchPromptSchedule node from ComfyUI_FizzNodes. The Scale and Effect multival inputs control motion amount and motion model influence independently. These inputs accept a single float, a list of floats, or masks, giving precise per-frame or per-region control over how strongly the motion module affects each part of the generation. ## SDXL, HotshotXL, and AnimateLCM Support Three additional motion module architectures are supported beyond the standard SD 1.5 modules. HotshotXL uses an SDXL checkpoint and requires the hsxl_temporal_layers.safetensors model. The README specifies that the context_length or total frame count should be 8 frames when not using context windows, and that the beta_schedule should be set to autoselect or linear (HotshotXL/default). AnimateDiff-SDXL is the official SDXL motion module from the AnimateDiff repository. The README describes it as remaining in beta after several months and notes that the beta_schedule should be autoselect or linear (AnimateDiff-SDXL). AnimateLCM allows faster sampling using the LCM sampler. The README specifies using autoselect, lcm, or lcm[100_ots] as the beta_schedule, applying the LCM LoRA, using the lcm sampler_name in KSampler nodes, and lowering cfg to roughly 1.0 to 2.0. ## Companion Node Packs The README lists several companion node packs that AnimateDiff workflows commonly use. ComfyUI-Advanced-ControlNet, also maintained by the same developer, makes ControlNets work with Context Options and allows specifying which latents in an animation sequence a ControlNet should affect. It includes SparseCtrl support. ComfyUI-VideoHelperSuite handles video loading, image-to-video assembly, and latent operations like appending, splitting, duplicating, selecting, and counting frames. comfyui_controlnet_aux provides ControlNet preprocessors not present in the base ComfyUI installation. ComfyUI_IPAdapter_plus adds IPAdapter support for guiding animation content from reference images. These are independent repositories with their own maintenance status. The README states that ComfyUI-Advanced-ControlNet and ComfyUI-VideoHelperSuite are maintained by the same developer as this extension, while the others are maintained by separate contributors. AnimateDiff workflows documented in the README and the documentation/ directory typically assume all or most of these are installed, so the effective setup involves installing several packages beyond this one. ## Limitations and the Right Baseline Expectations The motion modules themselves are not included in the repository and must be downloaded separately. Total download size depends on which modules are chosen; the README does not state file sizes, so planning storage ahead of time requires checking the source repositories. AnimateDiff-SDXL support is described as beta after several months, meaning SDXL animation workflows are more likely to require iteration and parameter tuning compared to SD 1.5. HotshotXL is limited to 8-frame context windows as its optimal setting. The extension requires ComfyUI 0.3.68 or greater, as specified in pyproject.toml. The [tool.comfy] section in pyproject.toml declares the publisher as kosinkadink and the display name as ComfyUI-AnimateDiff-Evolved, which is how it appears in the ComfyUI Registry at comfyregistry.org. That registry entry is the source that ComfyUI Manager searches when a user looks up the extension by name. The repository uses Apache-2.0 as its license. The last push was on 2026-07-28. The extension is for existing ComfyUI users, not a standalone animation tool. Generating video output requires ComfyUI-VideoHelperSuite or equivalent nodes to assemble frames into a video file. Workflow JSON examples are stored in the documentation/ directory and can be loaded directly into ComfyUI to reproduce documented results without rebuilding the node graph from scratch. The web/ directory contains any web extension files that modify the ComfyUI frontend interface for AnimateDiff-specific controls. ## Conclusion ComfyUI-AnimateDiff-Evolved is the practical choice for ComfyUI users who want to generate animations from image diffusion checkpoints using AnimateDiff motion modules. It works with most SD 1.5 motion modules out of the box and covers SDXL via HotshotXL and AnimateDiff-SDXL, though the README notes that SDXL support has remained in beta for several months. Before starting, confirm that the motion module matches the checkpoint architecture, that the beta_schedule is set to autoselect or the appropriate variant, and that at least 1 GB of storage is available for the motion module files. The last push was on 2026-07-28. ## FAQ ### Is there a tool called AnimateDiff for ComfyUI? Yes. ComfyUI-AnimateDiff-Evolved is a ComfyUI custom node pack by Kosinkadink that integrates AnimateDiff motion modules into ComfyUI's node graph. It is installed via ComfyUI Manager or by cloning into the custom_nodes directory. ### What does AnimateDiff do? AnimateDiff adds temporal motion to Stable Diffusion image generation by inserting a motion module that applies learned motion patterns across a sequence of frames. ComfyUI-AnimateDiff-Evolved extends this with sliding context windows for animations longer than the motion module's native frame count, plus Motion LoRA support, FreeInit, FreeNoise, and prompt travel. ### Can AnimateDiff be used to convert images to videos? AnimateDiff generates video sequences from diffusion checkpoints, and ComfyUI-AnimateDiff-Evolved supports img2vid workflows using ComfyUI's node graph. The README recommends ComfyUI-VideoHelperSuite for loading input images and assembling output frames into a video file. ## Sources - [Issues](https://github.com/Kosinkadink/ComfyUI-AnimateDiff-Evolved/issues) - [Kosinkadink/ComfyUI-AnimateDiff-Evolved on GitHub](https://github.com/Kosinkadink/ComfyUI-AnimateDiff-Evolved) - [License: Apache-2.0](https://github.com/Kosinkadink/ComfyUI-AnimateDiff-Evolved/blob/main/LICENSE) - [README](https://github.com/Kosinkadink/ComfyUI-AnimateDiff-Evolved/blob/main/README.md) --- Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/kosinkadink-comfyui-animatediff-evolved